Julia Sets and Their Relationship to the Mandelbrot Set

In complex analysis, elementary transcendental functions are extended to complex numbers through power series, revealing profound connections between seemingly unrelated functions: Euler's formula e^(iz) = cos(z) + i sin(z) unifies exponential and trigonometric functions, showing they are essentially the same function over the complex plane; the exponential function serves as a group homomorphism from complex addition to complex multiplication, with kernel 2πiℤ, making the complex logarithm multi-valued; complex powers z^w = e^(w log z) are generally multi-valued except when z is positive real or w is integer; hyperbolic functions relate to trigonometric functions via imaginary arguments (sin(iz) = i sinh(z), cos(iz) = cosh(z)); and complex analysis simplifies solving linear differential equations and Fourier series by treating them as algebraic operations on exponentials rather than separate trigonometric identities.

Holomorphic dynamics studies iteration of holomorphic maps on the complex plane, historically motivated by Newton's method for root-finding. The complex plane decomposes into two totally invariant sets: the Fatou set (stable orbits, normal family behavior) and the Julia set (chaotic boundary). Rational maps have infinity as regular point; transcendental maps have infinity as essential singularity, producing different Julia set structures. Fatou components classify as periodic, pre-periodic, or wandering. Fatou's 1926 landmark theorem established that for entire functions, periodic Fatou components can only be basins of attraction, parabolic basins, Baker domains, or elliptic components. This classification resolved major questions about possible dynamics within periodic Fatou components.

Rational functions P(z)/Q(z) are holomorphic except at zeros of Q(z). The exponential e^z is entire. Trigonometric functions are defined via Euler's formula: sin(z) = (e^(iz) - e^(-iz))/(2i), cos(z) = (e^(iz) + e^(-iz))/2, both entire. Hyperbolic functions sinh(z) = (e^z - e^(-z))/2, cosh(z) = (e^z + e^(-z))/2 satisfy the same identities as real functions. The complex exponential has period 2πi, meaning e^(z+2πi) = e^z.

This comprehensive section establishes the mathematical foundations of complex dynamics. It begins with basic definitions: analytic functions, iteration notation f^n, and the fundamental partition of the complex plane into Fatou set (nice dynamics) and Julia set (chaotic boundary). For polynomials, Bottcher's theorem provides conformal conjugacy near infinity, while Carathéodory's theorem links local connectivity of Julia sets to continuous Riemann map extensions—allowing complex dynamics to be understood via simpler angle-doubling maps. The section transitions to transcendental entire functions, introducing escaping sets replacing basins of infinity, and explaining how essential singularities fundamentally alter dynamical behavior. Singular values (critical and asymptotic) are defined, leading to class B classification requiring bounded singular sets. This foundational framework enables systematic comparison between polynomial and transcendental dynamics throughout the talk.

Transcendental functions are non-algebraic functions that include exponential functions (f(x) = b^x, where b is the base and x is the exponent) and trigonometric functions (which relate the sides and angles of right triangles). The six trigonometric functions are sine (opposite/hypotenuse), cosine (adjacent/hypotenuse), tangent (opposite/adjacent), cotangent (adjacent/opposite), secant (hypotenuse/adjacent), and cosecant (hypotenuse/opposite). The sine and cosine functions have a period of 2π, are defined for all real numbers, and have a range of [-1, 1]. Exponential functions are increasing when the base is greater than 1 and decreasing when the base is between 0 and 1.

Mandelbrot recognized that the finer indentations in the Koch curve were needed to model coastlines, leading to his famous article 'How Long Is the Coastline of Britain?' He argued coastlines are fractals and could measure roughness rather than length. Fractal dimension measures roughness - traditional geometry uses whole numbers (1D for lines, 2D for surfaces, 3D for solids), but fractals can have dimensions between whole numbers. The rougher a fractal is, the higher its fractal dimension. Julia sets, introduced by French mathematician Gaston Julia during World War I, involve iterating equations through feedback loops. Working by hand, you could never know what the complete set looked like, requiring fast computers to visualize.

This video presents an algorithm for estimating from below the Hausdorff dimension of Julia sets for holomorphic maps by constructing a MacMillan matrix whose entries are maxima of derivatives of pre-images of partition pieces, then defining Δ_k as the infimum of t for which the spectral radius of the matrix equals 1; this method has been applied to obtain new lower bounds for the Feigenbaum polynomial (greater than 1.4978), verify smoothness of Hausdorff dimension for certain parameters, and produce rigorous graphs of lower bounds for quadratic polynomials p_c(z) = z²+c with c∈[-2,2].

The Hausdorff dimension quantifies geometric complexity of fractal sets like Julia sets. For quadratic maps f_c(z) = z² + c, the dimension depends irregularly on c. Parameters where dimension equals 2 lie on the Mandelbrot set boundary, but for harmonic measure almost every boundary point, dimension is strictly less than 2. Yoccoz proved that for c ≠ 0, -2 in the Mandelbrot set, dimension exceeds 1, since connected Julia sets have dimension ≥1. At c=0 (unit circle) and c=-2 (interval), dimension equals exactly 1. Near c=-2, perturbations create complex Julia sets contained in ellipses with horizontal extent ~2√ε and vertical extent ~√(2ε). Joint work with Graczyk establishes that for c = -2 + ε, the Hausdorff dimension satisfies 1 + C√ε ≤ dim_H(J(f_c)) ≤ 1 + C√ε, showing dimension approaches 1 at rate √ε. Two main techniques estimate Hausdorff dimension: porosity (checking disks in complement at every scale) and beta numbers (measuring straight-line approximation). Parameter exclusion identifies large parameter sets with good dynamical properties. For c near -2, there exists a set A with -2 as a one-sided density point where disks near the critical value can be pulled back to large scales with bounded distortion. Exponential shrinking states that pre-images of large-scale balls shrink exponentially: diam(U) ≤ α^n for some α ∈ (0,1). Conformal measures satisfy μ(f(A)) = ∫|f'|^δ dμ, and for c near -2, conformal measures near the critical value scale as μ(B(0,R)) ≤ C·R^δ for R > ε^1000, with slightly worse estimates for smaller R.

Fractal dimension measures complexity or 'roughness' of fractals, with values between whole numbers (1 for lines, 2 for surfaces, 3 for volumes). Higher fractal dimension indicates greater roughness. Computers were essential for developing fractal mathematics because they could perform the infinite iterations required to generate fractal images. Julia sets are created by iteratively applying mathematical formulas to numbers, revealing complex fractal patterns impossible to visualize without computers. These mathematical structures demonstrate how simple rules can generate infinite complexity, a fundamental principle underlying fractal geometry.

Hausdorff dimension measures covering efficiency with varying square sizes; packing dimension uses uniform square sizes. Generally, Hausdorff ≤ packing dimension. Baker proved transcendental Julia sets always contain continua (dimension > 1). McMullen showed examples with Hausdorff dimension 2. Stallard proved all dimensions between 1 and 2 are achievable for Julia sets in the Eremenko-Lyubich class. With Simon Albrecht, the speaker constructed Julia sets with Hausdorff dimension approaching 1 using sequences of Eremenko-Lyubich functions with three singular values.

This video demonstrates how to render high-quality escape time fractals using a simple iterative algorithm where each pixel's coordinates serve as input to a complex number sequence; the rendering process involves iterating until the magnitude exceeds 2 or reaches a maximum iteration count, then mapping the iteration count to a gradient for coloring, with techniques like supersampling and multi-threading improving visual quality and performance.

This video presents two techniques for efficiently rendering Mandelbrot fractals by exploiting the set's topological properties. The rectangle processing algorithm evaluates escape times along rectangle boundaries and assigns these values to interior pixels when borders have uniform escape times, subdividing rectangles when escape times vary. The escape area processing algorithm traces escape time borders using a wall-hugging maze-solving algorithm, then fills interiors in one pass. Both techniques skip escape time evaluation for pixels inside processed regions, significantly reducing computational cost. However, pixelation of the Mandelbrot set can cause 'swallowing artifacts' where pixels are incorrectly assigned escape times from enclosing rectangles. These artifacts occur because the pixelated representation may appear disconnected even though the mathematical set is simply connected.

The Real Escape Time algorithm provides smooth color transitions in fractal visualization. When a sequence is close to the escape radius R but hasn't exceeded it, the next iteration z(n+1) ≈ z(n)^d (for z = z^d + c). The distance from the escape radius is approximately |z(n+1)| - R ≈ |z(n)|^d - R. To create smooth color transitions, the escape time is mapped to a continuous value using logarithms: first map the interval [R, R^d] to [1, D] using log base R, then map [1, D] to [0, 1] using log base D. This produces a continuous color gradient instead of discrete bands.

The escape algorithm determines set membership by checking if the modulus (magnitude) of the complex number exceeds 4. If the modulus is greater than 4, the point escapes and does not belong to the set. For rendering, the complex plane must be mapped to pixel coordinates: for a 640-pixel wide image, the real axis maps from -2 to 2, and the imaginary axis maps from -2 to 2. The mapping formula is: real = (x - width/2) * (4/width), imaginary = (y - height/2) * (4/height). This transforms pixel coordinates into complex plane coordinates, enabling visualization of the fractal.

This extensive section demonstrates practical implementation across three rendering approaches. Standard rendering tests one point per pixel, producing inaccurate blobby results with low iterations. Escape time coloring improves visibility by assigning colors based on escape speed, revealing more exterior structure but still missing fine detail. The distance estimation method emerges as superior by calculating gradients during iteration and using them to estimate proximity to the set. After escaping, the distance estimate |z| * log|z| / |dz| is mapped to grayscale using hyperbolic tangent. Side-by-side comparisons demonstrate dramatic improvements: distance estimation reveals extraordinary fine detail including intricate swirly patterns and connections between seemingly disconnected regions. The fundamental insight is that changing the question from 'is it in?' to 'how close is it?' unlocks previously inaccessible visual information, transforming our ability to explore fractal geometry.

The discrete Gompertz model exhibits complex dynamics: sigmoidal growth for α between 0-1, stable behavior for α between 1-2, bifurcation to two cycles at α=2, and chaotic behavior at α=2.5. Benjamin Gompertz developed this model in 1825 for mortality studies, now applied to epidemiology. Applied to COVID-19 in Spain, it projected peak cases around April 4 (200,000 infections) and deaths around mid-April (5,000). Epidemics follow sigmoidal, not exponential, growth patterns.

Chaos theory has been confirmed across diverse fields: fluid convection (Schaber's mercury experiments), visual perception (eyes responding to flashing lights), cardiac physiology (fibrillation progression), and dripping faucets. The Feigenbaum constant (4.669) describes the universal scaling of period-doubling bifurcations in any system with a single peak. This universality means chaos theory applies broadly, not just to specific equations. Scientists have even used chaos theory to control cardiac fibrillation by applying electrical shocks at precise moments, demonstrating practical applications of understanding complex systems.

The study of iterative functions zₙ₊₁ = zₙ² + c revealed profound connections between complex dynamics and fractal geometry. The Mandelbrot set classifies complex numbers c by whether sequences remain bounded, containing bulbs corresponding to periodic orbits of period 2ⁿ along the real axis. Julia sets fix c and vary initial conditions, with the Mandelbrot set representing origin points of all Julia sets. Mitchell Feigenbaum discovered universal constants governing period-doubling bifurcations, finding that the ratio of distances between successive bifurcation points converges to δ ≈ 4.669 as systems approach chaos. This critical exponent applies universally to any quadratic iterative function and extends to approximately quadratic unimodal functions, appearing in dripping faucets, thermal convection, and neuron firing—demonstrating that diverse natural phenomena share the same underlying mathematical structure near critical transitions.

Bifurcation is a scientific term describing a critical point in a system where small changes can lead to dramatically different outcomes. In complex systems, bifurcation represents a moment of instability where the system can either collapse into chaos or transition to a more stable, ordered state. The presenter explains that this concept, originally from physics and mathematics, applies to political and social systems where historical events can create irreversible turning points that determine future trajectories.

Bifurcation theory studies how small changes in parameters of differential equations can cause dramatic qualitative changes in system behavior, such as the emergence of limit cycles or chaotic attractors like the Lorenz attractor, where adding nonlinear terms to simple systems can transform them from predictable to highly complex dynamics through a sequence of bifurcations.